<div dir="ltr">In the packed representation we store not MPI_Datatypes but a handcrafted id for each one. The 2 codes should have been in sync. I'm looking at another issue right now, and I'll come back to this one right after.<div><br></div><div>Thanks for paying attention to the code.</div><div> George.</div></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Apr 30, 2015 at 3:13 PM, Satish Balay <span dir="ltr"><<a href="mailto:balay@mcs.anl.gov" target="_blank">balay@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Thanks for checking and getting a more appropriate fix in.<br>
<br>
I've just tried this out - and the PETSc test code runs fine with it.<br>
<br>
BTW: There is one inconsistancy in ompi/datatype/ompi_datatype_args.c<br>
[that I noticed] - that you might want to check.<br>
Perhaps the second line should be "(DC) * sizeof(MPI_Datatype)"?<br>
<br>
>>>>>>>>><br>
int length = sizeof(ompi_datatype_args_t) + (IC) * sizeof(int) + \<br>
(AC) * sizeof(OPAL_PTRDIFF_TYPE) + (DC) * sizeof(MPI_Datatype); \<br>
<br>
<br>
pArgs->total_pack_size = (4 + (IC)) * sizeof(int) + \<br>
(AC) * sizeof(OPAL_PTRDIFF_TYPE) + (DC) * sizeof(int); \<br>
<<<<<<<<<<<<br>
<br>
Satish<br>
<br>
<br>
On Thu, 30 Apr 2015, Matthew Knepley wrote:<br>
<span class=""><br>
> On Fri, May 1, 2015 at 4:55 AM, Jeff Squyres (jsquyres) <<a href="mailto:jsquyres@cisco.com">jsquyres@cisco.com</a>><br>
> wrote:<br>
><br>
> > Thank you!<br>
> ><br>
> > George reviewed your patch and adjusted it a bit. We applied it to master<br>
> > and it's pending to the release series (v1.8.x).<br>
> ><br>
><br>
</span>> Was this identified by IBM?<br>
><br>
><br>
> <a href="https://github.com/open-mpi/ompi/commit/015d3f56cf749ee5ad9ea4428d2f5da72f9bbe08" target="_blank">https://github.com/open-mpi/ompi/commit/015d3f56cf749ee5ad9ea4428d2f5da72f9bbe08</a><br>
><br>
> Matt<br>
<div><div class="h5">><br>
><br>
> > Would you mind testing a nightly master snapshot? It should be in<br>
> > tonight's build:<br>
> ><br>
> > <a href="http://www.open-mpi.org/nightly/master/" target="_blank">http://www.open-mpi.org/nightly/master/</a><br>
> ><br>
> ><br>
> ><br>
> > > On Apr 30, 2015, at 12:50 AM, Satish Balay <<a href="mailto:balay@mcs.anl.gov">balay@mcs.anl.gov</a>> wrote:<br>
> > ><br>
> > > OpenMPI developers,<br>
> > ><br>
> > > We've had issues (memory errors) with OpenMPI - and code in PETSc<br>
> > > library that uses MPI_Win_fence().<br>
> > ><br>
> > > Vagrind shows memory corruption deep inside OpenMPI function stack.<br>
> > ><br>
> > > I'm attaching a potential patch that appears to fix this issue for us.<br>
> > > [the corresponding valgrind trace is listed in the patch header]<br>
> > ><br>
> > > Perhaps there is a more appropriate fix for this memory corruption. Could<br>
> > > you check on this?<br>
> > ><br>
> > > [Sorry I don't have a pure MPI test code to demonstrate this error -<br>
> > > but a PETSc test example code consistantly reproduces this issue]<br>
> > ><br>
> > > Thanks,<br>
> > > Satish<openmpi-1.8.4.patch><br>
> ><br>
> ><br>
> > --<br>
> > Jeff Squyres<br>
> > <a href="mailto:jsquyres@cisco.com">jsquyres@cisco.com</a><br>
> > For corporate legal information go to:<br>
> > <a href="http://www.cisco.com/web/about/doing_business/legal/cri/" target="_blank">http://www.cisco.com/web/about/doing_business/legal/cri/</a><br>
> ><br>
> ><br>
><br>
><br>
><br>
<br>
_______________________________________________<br>
users mailing list<br>
<a href="mailto:users@open-mpi.org">users@open-mpi.org</a><br>
Subscription: <a href="http://www.open-mpi.org/mailman/listinfo.cgi/users" target="_blank">http://www.open-mpi.org/mailman/listinfo.cgi/users</a><br>
</div></div>Link to this post: <a href="http://www.open-mpi.org/community/lists/users/2015/04/26823.php" target="_blank">http://www.open-mpi.org/community/lists/users/2015/04/26823.php</a><br>
</blockquote></div><br></div>